Gradient-boosting threshold identification for developmental toxicity: A BPAF case-study framework for early-stage...

Polymers are essential to modern infrastructure, yet their dependence on persistent chemical additives continues to raise unresolved environmental and health challenges. Bisphenol AF (BPAF), a widely used substitute for Bisphenol A (BPA), has become a contaminant of concern due to evidence of developmental toxicity linked to its endocrine-active properties. Conventional risk assessments often rely on controlled laboratory or field-based data, limiting ecological relevance and early-stage evaluation accuracy. Given the limited availability of comparable toxicity datasets for most BPA analogues, this study applies BPAF as a case-study chemical to demonstrate a methodological framework for early-stage risk evaluation. We developed an integrated, multi-domain assessment framework that combines machine learning-derived toxicological benchmarks with spatial analytics to derive a screening-level reference dose (RfD) and evaluate environmental exposure patterns. Monitoring data show a pronounced imbalance in global monitoring efforts, with most data originating from Asia (n = 49) and Europe (n = 29). A conservative RfD-equivalent screening comparison identified several reported sediment and water-column concentrations in China and parts of Europe as preliminary risk flags, although these should not be interpreted as regulatory exceedances or formal water- or sediment-quality criteria. Spatial patterns and literature-reported source categories indicate that industrial and broader anthropogenic activities are the most frequently attributed sources of BPAF contamination. Priority monitoring should focus on under-reported regions, industrialized watersheds, and drinking-water sources. These findings underscore the need for strengthened chemical oversight and demonstrate the value of predictive toxicology and spatial analysis in supporting early-stage prioritisation and monitoring of BPAF.